- From: Paola Di Maio <paola.dimaio@gmail.com>
- Date: Wed, 10 Feb 2021 12:19:18 +0800
- To: W3C AIKR CG <public-aikr@w3.org>
- Message-ID: <CAMXe=SqhNQeeRRDrGVmtRW-hL3dyddsMoHzqGbiMTKLq3J1Log@mail.gmail.com>
We are (were) taught AI/ KR as either symbolic or subsymbolic/conenctionist and based on that neurosymbolism supposedly leverages both *Note, Its in the scholarly papers and textbooks, I did not suggest it myself, I am citing countless other authors* But the more we work with this subdivision the more with struggle with it and have to come to terms with the fact that it is notional, and possibly even misleading and wrong. Much discussion ensues. A quick post to share a great article, which I think points towards a more unified view of KR heading towards a more system view of KR Knowledge Representations in Technical Systems A Taxonomy Kristina Scharei, Florian Heidecker, Maarten Bieshaar https://arxiv.org/pdf/2001.04835.pdf i feel something can come from this view ----------------------------------------------------------------------------------------------------------------- Regarding the adequacy of KG as KR, (Owen you picked up on that in a reply to a post) well, adequacy can be an objective criterion in systems engineering for example, in AI systems modelling, we tackle bias an adequate KR should minimise or avoid bias altogether can KG alone, without other KR, prevent bias?
Received on Wednesday, 10 February 2021 04:20:13 UTC